5GMED Handover Test Results 

5G mobility can be divided into two main categories: beam level mobility and cell level mobility. The major difference between the two is that beam level mobility is handled by Layer1/PHY and Layer2/MAC signaling whereas cell level mobility requires some Layer3/RRC signaling. Cell level mobility can be further categorized into two types, namely idle mode mobility and connected mode mobility: 

  • The mobility mechanism in idle mode is called cell reselection. It is used by a UE in NR RRC Idle State or NR RRC Inactive state to camp on a better cell.  When cell reselection criteria are satisfied on a neighbor cell, the UE will autonomously (i.e. without any network involvement) camp on this new/better cell.  
  • The mobility procedure in connected mode is a Handover (HO). It can be defined as a mobility event happening when a UE in NR RRC Connected State is instructed by the network to change its current serving cell (source cell) to a neighbor cell (target cell) after the occurrence of a measurement event, i.e. when the neighbor cell becomes better than the current serving cell by a given offset.    

In the 5G MED project, the focus is clearly on connected mode mobility (i.e. with data traffic) rather than idle mode mobility (i.e. without data traffic).  Some drive tests were carried out at the Castelloli test track in Jan. 2023 and in the FR-SP corridor in Jul. and Oct.  The main goal of those mobility tests was to measure the three below HO-related KPIs: 

  1. The HO Preparation Time: the time elapsed between the Measurement Report Msg sent by the UE to the network and the RRC Reconfiguration (HO Command) Msg sent by the network to the UE 
  1. The HO Execution Time: the time interval between the RRC Reconfiguration Msg and the NR5G_RRC_HO_SUCCESS Event  
  1. The User-Plane HO Interruption Time: the time during which the UE receives no data from either the source cell or the target cell. 

To measure those HO metrics as well as some other Radio Access KPIs, we used the QXDM (Qualcomm Extensible Diagnostic Monitor) logging tool at the Castelloli test track and Keysight Nemo Outdoor drive test tool in the corridor. It is also worth noting that, due to different 5G NR frequency band deployments -n77 at Castelloli and n78 in the corridor-, we ended up testing two different devices: a Quectel RG520N-EU Evaluation Board and Vulcano-5G (Valeo TCU that embeds a Quectel AG551Q-EU module).  

Let’s discuss in further details the 5G HO procedures. As a matter of fact, as depicted in Figure 1, there are four main scenarios with regards to 5G Intra-RAT Handovers: Intra-gNB Handover, Inter-gNB Xn-Handover, Inter-gNB N2-Handover and Inter-gNB N14-Handover.  

Figure 1: Four Main 5G Intra-RAT Handover Scenarios 

As shown in Table 1 hereafter, the N2-based and N14-based HOs can be further expanded, depending on the PLMN configuration (Intra-PLMN or Inter-PLMN) and on the deployment of the SMF and UPF nodes:      

N2 and N14 HO sub-types Main Radio Access & Core Network entities involved in the HO  
Inter-gNB N2-Handover, Intra-PLMN, Intra-AMF, Intra-UPF S-gNB, T-gNB, AMF, UPF 
Inter-gNB N2-Handover, Intra-PLMN, Intra-AMF, Inter-UPF S-gNB, T-gNB, AMF, S-UPF, T-UPF 
Inter-gNB N14-Handover, Intra-PLMN, Inter-AMF, Intra-UPF S-gNB, T-gNB, S-AMF, T-AMF, UPF 
Inter-gNB N14-Handover, Intra-PLMN, Inter-AMF, Inter-UPF S-gNB, T-gNB, S-AMF, T-AMF, S-UPF, T-UPF 
Inter-gNB N14-Handover, Inter-PLMN, Home Routed Roaming H-gNB, V-gNB, H-AMF, V-AMF, H-UPF, V-UPF, H-SMF, V-SMF 
Inter-gNB N14-Handover, Inter-PLMN, Local Breakout Roaming H-gNB, V-gNB, H-AMF, V-AMF, V-UPF, V-SMF 

Table 1: N2 and N14 Handover sub-types  

As per 3GPP standards, there are two roaming scenarios supported by the 5G Core (discussed in our Cross-Border Roaming Challenges document): 

  • The Local Break Out (LBO) roaming scenario which makes use of the SMF and UPF located in the VPLMN and of the N14 control plane reference point for AMF-to-AMF communication. The user plane traffic is directly routed from the V-UPF to the Data Network (DN).    

Figure 2: Inter-PLMN Handover in the case of LBO roaming scenario 

  • The Home Routed (HR) roaming scenario which makes use of the SMFs and UPFs present in both the VPLMN and the HPLMN. The data traffic is redirected to the HPLMN core network via the N9 user plane reference point and SMF-to-SMF communication is done through the N16 control plane reference point. 

Figure 3: Inter-PLMN Handover in the case of HR roaming scenario

The HO metrics collected during the Jan, Jul & Oct 2023 drive test sessions are shown in the below table (Table 2), with all the time periods/durations expressed in milliseconds: 

Table 2: Handover metrics

Here after are some comments & observations about those test results:  

  • Back in January 2023, intra-gNB Handover was the only mobility type being supported – and thus tested – at the Castelloli test track. In the corridor, we tested Intra-gNB HO, Inter-gNB N2 HO and Inter-gNB Inter-PLMN N14 HO. 
  • Inter-PLMN HO between French gNB PCI#553 (NR-ARFCN 642732, Nokia Vendor) and Spanish gNB PCI#26 (NR-ARFCN 647394, Ericsson Vendor) was still not working late August 2023. Considering that a demo of 5GMED was planned to be given to the European Commission late October, the project management decided to move the location of the Inter-PLMN handover more South into Spanish territory, thus eliminating the difficulties of an inter-freq and inter-vendor setup. Doing so led to a successful demonstration of an Inter-PLMN HO in HR mode, between French gNB PCI#25 (NR-ARFCN 647394, Ericsson Vendor) and Spanish gNB PCI#192 (NR-ARFCN 647394, Ericsson Vendor). 
  • The lowest HO preparation times were measured at Castelloli.  
  • It is unclear to us why the HO preparation times measured in the Corridor during the October test session are so widely spread out.      
  • The duration of the HO preparation phase depends more on the radio and core networks than on the UE/TCU.  Here are some sources of latency during the HO preparation phase: gNB processing times (processing of the UL L3 RRC Measurement Report Msg, HO decision at the source cell, admission control at the target cell), N2 interface delays, processing of the N2 messages, preparation and transmission of the DL L3 RRC Reconfiguration Msg. 
  • As shown in Table 3 hereafter, the HO execution time is made up of two other components: the HO-to-Preamble Time and the RACH access delay. The HO-to-Preamble time is the time elapsed between the L3 RRC Reconfiguration Msg and the triggering of the L2 MAC Random Access (RA) procedure. The RACH access delay is the delay of the L2 MAC RA procedure, from triggering till completion. 

Table 3: Breakdown of the HO execution time  

  • The variations in the HO execution times are caused by the variations in the RACH access delays. 
  • We can see that the RACH access delay accounts for about 20 to 35% of the HO execution time, which is quite significant.  
  • The user plane HO interruption times and the HO-to-preamble times measured at the Castelloli test track were very stable, ~35ms and ~39ms respectively. In the corridor, if the HO-to-preamble times were also quite stable, we can observe a higher fluctuation in user plane HO interruption times. 
  • The type of Random Access procedure used by the two radio access networks for HO purpose is different: 4-step CFRA (Contention-Free Random Access) in the French network and 4-step CBRA (Contention-Based Random Access) in the Spanish network. We were expecting the CFRA procedure to be quicker than the CBRA procedure, but the test results don’t really show that.  

These handover results will be further expanded with tests from the different use cases presented in the project.